SILO II is one of the smart application modules used in large-scale industrial processes for advanced control and optimization. 

The main application of the SILO II system is combustion process optimization. Heat and electricity producers can minimize NOX, CO and SO2 emissions and increase generation efficiency. 

SILO II also can be used to increase plant controllability. Therefore, costs related to the power generation process itself, and to emissions control regulations are reduced. Furthermore, when using SILO II, companies may avoid the higher costs of emission mitigation systems allowing for substantial increases in their infrastructure efficiency. SILO II could also be used for FGD, SCR and SNCR optimization.

SILO II system performs an on-line optimization of MIMO (Multi Input Multi Output) industrial processes. It is responsible for:

• Maximization of income that is related with output product,
• Minimization of costs which are related with fuel costs and penalties for air pollution.

SILO II Graphical User Interface is based on WWW technology. All users which are connected to intranet can have access to the SILO II interface.

The task of SILO II is to perform an on-line optimization of the current process operating point. The optimizer is implemented above the base control layer in a layered control structure (refer fig. 5). 

The SILO II system calculates setpoints or setpoints corrections for controllers that operate in the base control layer. Control systems of power units in power plants are based on PI (Proportional-Integral) controllers. 

These controllers control sub-processes (e.g. oxygen level in exhaust gases or windbox to furnace differential pressure in the case of a power unit control system) that have an influence on a main optimized MIMO (Multi Input Multi Output) process (e.g. combustion process in the case of a power unit control system).

In our description the following terminology will be used:

• y – main process outputs vector (combustion process example: NOX and CO emission, steam temperatures, output gases temperature, loss of ignition and other factors related with unit efficiency),
• d – main process disturbances vector (combustion process example: unit load, coal mill configuration, measured or estimated fuel calorific value, etc.),
• md – decision vector - setpoints for low level controllers (combustion process example: oxygen setpoint, auxiliary air dampers positions demand, OFA dampers positions demand, OFA tilts demand, burner tilts demand, coal feeder demand, FD and ID fan balance corrections, correction of wind box to furnace differential pressure setpoint, etc.),

• mt – traced setpoints for low level controllers (combustion process example: oxygen setpoint that enters the PI controller in DCS system, demand signal that is sent to OFA dampers, etc.),
• mc – vector of measured sub-process outputs - inputs to the main optimized process (combustion process example: measured oxygen content in flue gases, measured OFA position, etc.),
• mf – control variables availability vector (combustion process example: status of the oxygen control loop - SILO II supervisory enabled/disabled, etc.),
• mp – operator setpoints vector (combustion process example: operator demand for oxygen, etc.).

SILO II is a completely new solution for industrial process optimization. This new approach is based on analogy with the immune systems of living creatures. Other advanced process control and optimization solutions are based on a large, sophisticated models, that are difficult to obtain, maintain and tune. SILO II uses a different approach. It gathers portions of knowledge about the process, and uses selected portions of knowledge to forecast plant behavior in a close neighborhood of the current process operating point. 

SILO II learns static process responses for a given control change. In particular it stores packets of information which are used by the optimization module to create a direct model of inner process dependencies in the neighborhood of the current process operating point. Moreover, SILO II is able to gather knowledge about the process, based on historical data that may be stored within a DCS system.

Using its own knowledge base, SILO II is able to identify dependences between process inputs and outputs for each analyzed process operating point. It is a brand new approach in process non-linearity handling. SILO II is also able to automatically create a mathematical model in every optimization period (e.g. every 5 minutes). It assures more direct adaptation to a non-linear characteristic in comparison with other solutions, which use manually created models.

There is no need to perform long lasting and labor consuming identification experiments. SILO II learns the process in on-line mode and increases its efficiency over time (please refer fig. 2). At the beginning, after a SILO II installation, it has no knowledge about the process. SILO II uses a special heuristic to optimize the process and learns the basic process dependences. After a few hours it can start to create models. The accuracy of these models continually improves over time so that after one week, the solution can perform an efficient optimization of a process. 

Key benefits for the user:

• There is no need to change the production schedule of a plant with long-lasting parametric tests. Thus avoiding significant loses for changes in production plan.
• There are no significant inefficiencies related to process operation with parametric tests. Long-lasting parametric tests can cause significant loses for a plant because process operation was dictated by testing requirements. This is also avoided with SILO II.
• The cost of implementation is significantly reduced, so there is a higher return of income for the customer.
• SILO II significantly reduces customer loses related with manual model creation process.

There is also no need for additional SILO II tuning, thanks to a very efficient adaptation mechanism. Over time SILO II can compensate for poor performance of the base control structure that is related to poor adaptation. It reduces the customer cost related with some base control structures tuning.

SILO II is dedicated for different types of large-scale plants. In case of combustion process optimization, it can handle all types of energetic boilers, different fuels and different load profiles. SILO II reaches highest performance during steady-state control, however there is a special dedicated mechanism that is activated during a transition state. Thanks to this mechanism, SILO II is able to handle operating point transitions (e.g. load changes). The SILO II optimizer is very easy to install and maintain. It has an efficient adaptation mechanism and a flexible structure of the optimization task, thus it can follow changes of process characteristics and modifications of production strategy. All these features make SILO II a perfect solution for a fleet optimization approach.

SILO II can be integrated with ZOLO BOSS measurements. The ZOLO BOSS is a new technology of measurement. It uses tunable diode laser absorption spectroscopy to measure temperature, O2, CO, CO2, H2O across the furnace. The ZOLO BOSS consists of set of lasers placed at different furnace elevations. A set of ZOLO BOSS lasers from a single elevation creates a grid representing two-dimensional map of combustion process parameters. SILO II is able to read ZOLO BOSS measurements and use them to control a 3D shape of the fireball. SILO II controls the fireball’s parameters such as: horizontal and vertical shift, angle, intensity and concentration, in order to optimize efficiency and emissions.

The newest version of SILO II has a new algorithm that is able to handle a significant process transition in an effective way. This algorithm is implemented in the Transition State layer in the Optimization module of SILO II. In the case of a combustion process optimization, this new mechanism allows for optimization of relatively small power units characterized by frequent transitions of unit load.

 The AIT (Automatically Identified Targets) and UDT (User Defined Targets) are used to move the control vector value to a point that lies in a close neighborhood of an optimal solution related with the new process operating point. This transition is fast. A new process operating point is a starting point for model based optimization.